Soft Matter

PAPER

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article OnlineView Journal | View Issue

aAix Marseille Universite, CNRS, Central

Marseille, France. E-mail: deschamps@irph

Fax: +33 413552001; Tel: +33 413552083bAix Marseille Universite, CNRS, CINaM UMcFaculte de Pharmacie, Universite de Reims

7312, 51687 Reims, France

† Electronic supplementary informationunload of the AFM probe during one inde

Cite this: Soft Matter, 2014, 10, 4561

Received 13th February 2014Accepted 3rd April 2014

DOI: 10.1039/c4sm00349g

www.rsc.org/softmatter

This journal is © The Royal Society of C

Mechanical characterization of cross-linked serumalbumin microcapsules†

Clement de Loubens,a Julien Deschamps,*a Marc Georgelin,a Anne Charrier,b

Florence Edwards-Levyc and Marc Leonetti*a

Controlling the deformation of microcapsules and capsules is essential in numerous biomedical

applications. The mechanical properties of the membrane of microcapsules made of cross-linked human

serum albumin (HSA) are revealed by two complementary experiments in the linear elastic regime. The

first provides the surfacic shear elastic modulus Gs by the study of small deformations of a single capsule

trapped in an elongational flow: Gs varies from 0.002 to 5 N m�1. The second gives the volumic Young's

modulus E of the membrane by shallow and local indentations of the membrane with an AFM probe: E

varies from 20 kPa to 1 MPa. The surfacic and volumic elastic moduli increase with the size of the

capsule up to three orders of magnitude and with the protein concentration of the membrane. The

membrane thickness is evaluated from these two membrane mechanical characteristics and increases

with the size and the initial HSA concentration from 2 to 20 mm.

1 Introduction

Microencapsulation refers to diverse techniques to encloseactive materials within a shell with the aim of protecting themfrom the outside and to control their spatiotemporal release.This process offers answers to many biotechnological chal-lenges1–3 such as cancer therapy4 and cardiovascular treat-ments.5 Various containers result from encapsulation of adroplet coated with a solid, such as polymeric capsules, orliquid membranes such as vesicles. The membrane may exhibitvarious mechanical properties that are essential for controllingthe delivery of the active materials.6–8 These characteristics arequite limited for uid vesicles made of lipids:9 the thickness isxed by the lipid bilayer and their deformation is governed bybending rigidity and membrane incompressibility.10 While themembrane viscosity is negligible for vesicles, polymersomes arealso characterized by shear resistance.11 The variety of geomet-rical and mechanical properties is widely increased for capsulesmade of polymers with weak or strong cross-linking. Themembrane is supposed to exhibit a viscoelastic behavior and abending resistance. These characteristics depend on both thechemical composition of the membrane and the preparationprocess. Understanding the role of the process on the

e Marseille, IRPHE UMR 7342, 13384

e.univ-mrs.fr; leonetti@irphe.univ-mrs.fr;

R 7325, 13288 Marseille, France

Champagne-Ardenne, CNRS, ICMR UMR

(ESI) available: Fig. S1 of the load andntation. See DOI: 10.1039/c4sm00349g

hemistry 2014

mechanical properties of the membrane is thus of primeimportance.

Various experiments12,13 have been developed to test themembrane mechanical properties of capsules. The rst methodis dedicated to local stresses applied to the capsule. The prin-ciple is to put a probe in contact with the membrane to study:the compression between two plates,14 the AFM scanning with asharp tip15 or a large colloidal particle12,16 and the micropipetteaspiration.17 The second method is devoted to global stressesapplied to the capsule by means of hydrodynamic ows to studythe capsule deformation in a spinning drop apparatus,18 insidea capillary19 or in a shear ow.20,21 Several of these techniquesare based on theoretical studies,22 which are also useful tovalidate numerical studies.23–26

In the elastic regime (i.e. under small deformations), thecapsule behavior under hydrodynamic stresses has beenexplored theoretically,22 giving a relationship between thedeformation of the capsule and the surfacic shear modulus. Therst experiments which conrmed these predictions were con-ducted in shear ow27 based on a cylindrical Couette device andin an elongational ow generated with the so-called four rollmill apparatus.20

In this paper, we investigated the mechanical behavior ofmicrocapsules that were manufactured by interfacial cross-linking28,29 of human serum albumin (HSA) with terephthaloylchloride in a water-in-oil emulsion system. Thesemicrocapsulespresent the advantage of having a biocompatible, biodegrad-able and stable membrane for medical applications. The prep-aration process allowed us to vary easily the HSA concentration.The size distribution of the microcapsules in the differentbatches was also dependent on the preparation process (Fig. 1).

Soft Matter, 2014, 10, 4561–4568 | 4561

Fig. 1 Size distribution of each batch of capsules: 5% HSA (solid line),10% (dashed line) and 20% (dash-dotted line).

Soft Matter Paper

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article Online

Two complementary experimental methods were used todetermine surfacic and volumic mechanical properties of themembrane in the elastic regime. Studying the deformation ofthe whole capsule in an elongational (or hyperbolic) owallowed us to obtain the surfacic shear modulus Gs of themembrane, whereas its volumic Young's modulus E wasmeasured by local and small deformations with an AFM tip. Thecombination of these surfacic and volumic moduli led to thecalculation of the membrane thickness.

Fig. 2 Left: an image of channel 1. The cross-like channel width is 1mm. Arrows show the flow direction. The circle defines the regionwhere the elongational rate variation does not exceed 5%. Right: anexample of velocity field obtained within the measurement area withPTV.

2 Materials and methods2.1 Capsule preparation

Batches of cross-linked HSAmicrocapsules were prepared usingthe interfacial cross-linking method,28,29 with terephthaloylchloride as the cross-linker. HSA was provided by LFB Bio-medicaments, as a 200 mg mL�1 solution. This solution hasbeen freeze-dried to vary the HSA concentration for the micro-encapsulation process. The organic solvents (chloroform andcyclohexane), cross-linking agent (terephtaloyl chloride) andsurfactants (sorbitan trioleate and polysorbate) were purchasedfrom SDF, Acros Organics and Sigma, respectively.

Some of the preparation parameters were varied in order toobtain membranes with different cross-linking degreessurrounding liquid droplets of various sizes. Briey, the HSAsolution was prepared at various concentrations (5%m V�1, 10%,20%) in a pH 8 phosphate buffer. This aqueous solution wasemulsied in cyclohexane containing 2% (w/v) sorbitan trioleateat a stirring speed of 625 rpm. A 2.5% (w/v) solution of tereph-thaloyl chloride in chloroform : cyclohexane (1 : 4 v/v) was thenadded to the emulsion and the cross-linking reaction was allowedto develop for 30 min. The reaction was stopped by dilution of thereaction medium. The microcapsules were separated from theorganic phase by centrifugation and washed successively withcyclohexane, and with water containing 2% (w/v) polysorbate, andnally were transferred into pure water. The granulometricdistribution of each batch was determined by laser diffractionusing a Malvern Mastersizer 2000 (Fig. 1). In order to be usedeither in elongational ow or in AFM apparatus, capsules from

4562 | Soft Matter, 2014, 10, 4561–4568

original batches were diluted to about thousand times and mixedvery gently for at least 24 hours in glycerol of 98% of purity (VWR).

2.2 Elongational ow

An elongational ow was generated by using a cross-likechannel made of two PMMA plates sealed together. One of thetwo was rst countersunk to create the uid path and allowimage capture (Fig. 2). Two different channels were used withsquare cross-sections of 1 and 4 mm2 named thereaer channel1 and channel 2, respectively. The uid was injected into thechannel through a glass syringe mounted on a home-madesyringe pump based on a PI actuator M235-52S. The injectionspot was unique. The channel was then split in two symmetricalways, which were recombined at the stagnation point. Twocontainers at the ambient pressure were plugged to the twooutlets to enable the complete stop of the ow at rest. We paidattention to avoid any deformable pipe and any air bubbleswithin the whole uidic system in order to minimize the tran-sient time when the ow was switched on. The visualization wasachieved with an inverted microscope Olympus IX-71 withmagnication 10 to 16�. A high speed video camera PhotronFastcam SA3 allowed us to acquire up to 5000 frames persecond. The image post-processing was performed with Matlab.

Elongational rates were measured by Particle Tracking Veloc-imetry (PTV). Spherical polystyrene particles of 10 mm diameter(Bangs Laboratories) were seeded in glycerol. The observationarea was 1.1 � 1.1 mm2 seen at 16� in channel 1 and 1.7 �1.7 mm2 at 10� in channel 2. The velocity eld of the ow was asfollows in the plane (x, y) (Fig. 2): ux ¼ _3x, uy ¼ �_3y with _3 theelongational rate. We dened a circular area centered at thestagnation point of the ow for which _3 was considered homo-geneous, i.e. with a maximal standard deviation of 5%. Forchannel 1, the measurement region was of 520 mm diameter. Forchannel 2, which had its corners cut in the cross-like region, themeasurement area was of 1300 mm diameter. _3 was averaged overthese regions. The deformation of the capsules was analyzedexclusively in these regions. The ow rate was varied from 10 to500 mL s�1 to measure the elongational rate at the middle of thethickness of the channel. We found a linear dependence of theelongational rate on the ow rate (Fig. 3) for both channels.

This journal is © The Royal Society of Chemistry 2014

Paper Soft Matter

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article Online

The dependence of _3 on the depth of the channel wasdetermined (z direction). The ow rate was kept constant at 17mL s�1. _3 was measured through the thickness of the channel todetermine its variation with the depth z. As expected, the owhad a Poiseuille-like prole (Fig. 4). For the measurement of theshape of capsules in the elongational ow, its z position wasmeasured to determine accurately the elongation rate using thelaw _3 ¼ f(z) determined in Fig. 4. To prevent any effect of thechannel walls, only capsules close to the center (i.e. distancefrom the center below � 200 mm) were studied.

To deduce the stress h_3, the viscosity h of glycerol wasmeasured with a rheometer ThermoScientic Haake Mars III ina cone–plate geometry (2� cone, 60 mm diameter) in order to geta correction of the viscosity with the temperature. Thetemperature of the sample was regulated by an integratedPeltier effect system that heated the plate of the cone–platemeasurement conguration. The regulation system ensuredthat the temperature of the lower plate was accurate to within�0.1 �C. The measurement geometry was enclosed in an enve-lope that acts as a solvent trap, thus considerably reducingevaporation on the unconned surface of the sample. It alsoreduced heat losses and ensured a uniform temperature aroundthe sample. Calibration with a standard oil (Paragon Scientic

Fig. 3 Elongational rate as a function of the flow rate for channel 2.Only the upper half is plotted.

Fig. 4 Elongational rate as a function of the elevation z for channel 2.

This journal is © The Royal Society of Chemistry 2014

Ltd) of viscosity similar to that of glycerol showed that thestandard error on the viscosity measurement was below �2.5%.A digital thermometer, that was calibrated with the Peltiersystem of the rheometer, was inserted very close to the channelto get the temperature with a precision of 0.1 �C. All experi-ments were carried out at a temperature range of 22 �C� 0.5 �C.

The deformation of an elastic capsule under ow can beevaluated by the dimensionless ratio of the viscous stress h_3 onthe elastic response Gs/R. This is the capillary number Ca:

Ca ¼ h3:R

Gs

(1)

where R is the radius of the capsule in the resting state (Fig. 5), his the viscosity of the outer uid and _3 is the rate of deformationdue to a shear or elongational ow. Various constitutive lawsmay describe the linear and non-linear behaviour of a cross-linked polymer membrane such as Neo-Hookean, Skalak orMooney Rivlin.30 However, all these models reduce to Hooke'slaw in the limit of small deformations and thus leading to thesame mechanical moduli. The membrane mechanics of a thin3D isotropic material can be modelled by a surfacic constitutivelaw without bending resistance and with the following rela-tionship between the relevant mechanical quantities:

Gs ¼ Es

2ð1þ nÞ (2)

Es ¼ Eh (3)

where Gs is the surfacic shear elastic modulus, n is the Poissonratio, Es and E are the 2D and 3D Young's moduli and h is thethickness of the membrane. The Poisson ratio n was assumed tobe equal to 1/2 (i.e. the membrane is considered as an incom-pressible material).

The projection in the plane (x, y) of the deformed capsulewas assumed to be an ellipse of major and minor semi-axislengths L and B. In the small deformation regime, the defor-mation is evaluated by the shape parameter (Fig. 5):

D ¼ L� B

Lþ B

Capsules were trapped one by one at the stagnation point ofthe elongational ow (Fig. 1). This was achieved by regulating

Fig. 5 Image of a capsule of radius R at rest (left). On the right thesame capsule deformed by the elongational (or hyperbolic) flow. L andB are the major and minor semi-axis respectively. Notice that thedeformation is large (14%) compared to what has been realized in thepaper. Scale bar is 100 mm.

Soft Matter, 2014, 10, 4561–4568 | 4563

Soft Matter Paper

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article Online

the hydrostatic pressure at the outlets and by moving back andforth the piston of the syringe as oen as necessary. When thecapsule was stabilized, we applied suddenly the ow andmeasured the deformation D of the capsule as a function oftime. D saturated to a stationary value called DN (Fig. 6). Themeasurement area was large enough for the capsule to reach astationary deformation before it was carried away by the ow.For each capsule, we repeated several times the same experi-ment at different elongational rates by bringing the capsulebackward to the stagnation point. Finally, it has been showntheoretically that DN varies linearly with the capillary numberDN ¼ 25/6 Ca.30 Thus using eqn (1), the determination of Gs isgiven by:

Gs ¼ 25

6

hR3:

DN

(4)

2.3 AFM

Force measurements were carried out in glycerol using AFM(NTEGRA from NT-MDT). sQube colloidal probes with a silicondioxide bead and the resonance frequency in the range 6–21kHz were purchased from NanoAndMore. Probe tip radii weredetermined aer imaging the tip by scanning electron micros-copy and a radius of 0.95 mm was found for every tip. Springconstants ranging from 0.03 to 0.13 N m�1 were determinedusing the thermal noise method aer obtaining the deectionsensitivity of the cantilever by pressing the AFM tip against ahard reference glass bead of the 70 GPa Young's modulus.Cantilever deection sensitivity measurements were performedbefore all sets of measurements. Force measurements wererealized at a loading rate of 1600 nm s�1 and the indentationdepth was maintained below 150 nm for most samples. Foreach capsule, statistical analysis was realized with a minimumof 200 force measurements obtained from a total area of 15 �15 mm2. To extract the Young's modulus, each curve was ttedusing the Hertz model (see ESI† for the tting model) of contactbetween two spheres according to the following equation:

Fig. 6 Deformation parameter D for one capsule (20% HSA) and fourelongational rates: 3.53 s�1, 5.30 s�1, 7.13 s�1, and 9.04 s�1. DN wasextracted as the plateau of each curve and is emphasized here by thedotted lines.

4564 | Soft Matter, 2014, 10, 4561–4568

F ¼ 4

3

~R1=2

d3=2�1� nt2

��Et þ

�1� nc2

��Ec

with F being the force applied by the tip on the capsule, Et¼ 150GPa and E the Young's moduli of the tip and the capsule andnt ¼ n ¼ 0.5 the Poisson ratios of the tip and the capsulerespectively. ~R ¼ (1/R + 1/Rt)

�1 with R and Rt being the capsuleand tip radii respectively, and d the indentation. As veriedexperimentally in this range of indentation, this modelassuming an elastic behavior of the capsules is valid.

3 Results and discussion3.1 Surfacic shear modulus

Fig. 6 shows the typical deformation curves as a function of timefor one capsule (20% HSA) and four elongational rates. Thestationary deformation DN was dened as the plateau value ofthe curve D(t) at steady state. In Fig. 7, we report this maximaldeformation as a function of the hydrodynamic stress h_3. Aspredicted from theory,22 small deformations increase linearlywith the stress. All the capsules were observed within the linearelastic regime, typically D < 6% in this paper. In fact, all theindividual curves DN as a function of the stress were tted witha linear regression. Typically, the reliability factor values variedfrom 0.92 to 1. Moreover, some capsules were not perfectlyspherical and could be characterized by an initial deformationD0 in the resting state. Only capsules with D0 < 0.2% wereanalyzed. Aer the hydrodynamic stress was applied, thecapsule recovered its initial shape, i.e. its radius R and its initialdeformation D0 did not vary. Thus, experiments were carriedout in the elastic regime.

Each capsule was tested for about four to six different valuesof elongational rates covering a range with at least a factor 2.The surfacic shear modulus Gs for each run was calculated byeqn (4) and then averaged over a full set of runs for eachcapsule. Fig. 8 shows the variation of Gs with the capsule radiusR for 10% HSA. Notice that measurements are independent ofthe size of the channel.

Fig. 7 The deformation DN as a function of the stress for the capsule(20% HSA) shown in Fig. 6. The linear regression depicts the linearelastic regime.

This journal is © The Royal Society of Chemistry 2014

Paper Soft Matter

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article Online

The results of Gs for the three HSA concentrations (5, 10 and20%) are shown in Fig. 9. As the size of the capsule increases,the surfacic shear modulus strongly non-linearly rises. For 10%HSA, a factor 3.5 in size gives a factor 2000 in Gs. Small capsules(about 70 mm radius) exhibit a Gs of the order of 10�2 N m�1

whereas a larger one (about 160 mm radius) is characterized by aGs of about 5 N m�1. Compared to the size distribution of thecapsules for each HSA concentration (Fig. 1), the measurementsof Gs were limited to capsules whose radius was less than 200mm. In fact, the experimental area where measurements arevalid (see Fig. 2) was too small to observe steady deformationsfor radii larger than 200 mm. Furthermore capsules with 10 and20% HSA larger than 170 mm were too rigid to be sufficientlyelongated with our experimental set-up. Capsules with 5% HSAsmaller than 100 mm were non-spherical at rest, so that no dataare reported in the gure. The method consisting in subtractingthe initial deformation D0 to D18 was not applicable here sincethe initial deformation is not systematically elliptic.

Fig. 8 Surface shear modulus for capsules with 10% HSA as a functionof the radius R. Open squares: capsules in channel 1. Filled squares:capsules in channel 2.

Fig. 9 Surfacic shear modulus as a function of the radius R and therate of HSA. Lines are the interpolation of data points with the functionlog(Gs)¼ axb + c. See values of (a, b, c) for each batch below in the text.Circles: 20% HSA capsules. Squares: 10% HSA capsules. Triangles: 5%HSA capsules. The error bar on each point is less than 6%, it means lessthan the size of the three symbols.

This journal is © The Royal Society of Chemistry 2014

The large range of sizes accessible with the present methodand its resolution allowed us to measure Gs in the elastic regime(DN < 6%) and to show its variation with capsule size and HSAconcentration. None of the former studies on other kinds ofcapsules have mentioned the variation of Gs according to thesize.20 It was even avoided for capsules based on the poly-siloxane network18 for which attention was paid to keep themembrane constitution equivalent regardless of the size. Ourresults show (Fig. 8) that the effect of the size on Gs is lessimportant for large capsules than for small capsules. Thus,measuring Gs with an elongational ow offers both a very smalldegree of dispersion of the measurements and a method toinvestigate accurately the dependence of Gs with the processparameters. However, capillary experiments19 have to bepreferred for large and rigid capsules since it is easy for applyinghigher stresses.

The shear modulus is much dependent on the HSAconcentration. In Table 1, Gs values are reported for radii about107 mm. As the quantity of proteins increases in the emulsi-cation preparation by a factor 4, the resistance of the membraneto shear is multiplied by 30. This behavior is consistent withother observations on capsules prepared with variousmaterials.18,21,31

The elongational method is peculiarly accurate to measurethe surfacic shear modulus Gs characterized by an error lessthan 6% for each capsule. It means that the deviations largerthan this error result only from the capsule processing. Such adependence was expected as the processing is complex andinvolves some chemical reactions, a biphasic system, diffusionof molecules and an hydrodynamic ow, the interfacial poly-merization taking place during the solution mixing. As shownin Fig. 9, the deviation increases with the HSA concentrationand is larger for the batch 20% and smaller for the batch 10%.

3.2 Volumic Young's modulus

Similar to the surfacic shear elastic modulus Gs, the volumicYoung's modulus E of the capsule membrane varies with thesize R of the capsules (Fig. 10) whatever be the HSA concen-tration. Due to the difficulty of these measurements, experi-mental data do not cover the same range of size for each batchthan the previous hydrodynamic study. One limiting step isgiven by the lubrication phenomenon in the thin glycerol lmbetween the substrate and the capsule governing the time toreach the stationary state. During an indentation of depth dwitha maximum of the order of 100 nm, the zone of deformationextends to the characteristic length lz

ffiffiffiffiffidr

pwhere r ¼ 0.95 mm

is the tip size of the AFM probe: l z 0.3 mm. This order ofmagnitude is sufficiently small compared to the capsule size

Table 1 Variation of Gs for radii about 107 mm

Capsules R (mm) Gs (N m�1)

5% HSA 107.0 0.078 � 0.002310% HSA 105.9 0.669 � 0.01420% HSA 108.4 2.29 � 0.095

Soft Matter, 2014, 10, 4561–4568 | 4565

Fig. 10 Young's modulus E of the membrane measured with AFM forthe different HSA concentrations. Circles: 20% HSA capsules. Squares:10% HSA capsules. Triangles: 5% HSA capsules.

Soft Matter Paper

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article Online

and the expected membrane thickness to insure the relevanceof the Hertz law (see Materials and methods).

AFM experiments are local while deformations of a capsulein an elongation ow depend on the global shape. Thus thenon-sphericity of the capsule does not affect the measurementas it is the case for capsules with 5% HSA smaller than 100 mmin the elongational ow.

The smallest observed capsule leads to a Young's modulus ofabout 0.02 MPa whereas for the largest one, E is 100 timeslarger. Considering the HSA concentration, we qualitativelyobserve the same behavior as for Gs: E increases by one order ofmagnitude as the HSA rate increases from 5 to 20%. However,this dependence is weaker for batches 10 and 20% HSA. Thisindicates that the properties of the polymer network saturatewith an increase of the HSA concentration. Indeed, batches 10and 20% HSA give the same E and the same Gs for capsules of170 mm radius which emphasizes that both E and Gs arecorrelated. Contrarily, the Young's modulus increases clearlywith the capsule radius for 5% HSA. No variation of E with thesize of the capsule has been reported elsewhere even forcapsules made of other kinds of albumin such as ovalbumin.

Fig. 11 The calculated thickness h of the capsule membrane as afunction of the size R. Circles: 20% HSA capsules. Squares: 10% HSAcapsules. Triangles: 5% HSA capsules. The dashed lines are a guide tothe eye.

3.3 Membrane thickness

In this part, it is shown that the two previous measurements bytwo complementary experiments allow us to infer the thicknessof the capsule membrane. As the data of Young's modulus aresparse, only orders of magnitude can be evaluated.

So far the membrane has been considered as a 2D mediumbut due to the interfacial cross-linking reaction, the membranehas a nite thickness h. It was assumed in previous studies oncapsules made of albumin that the thickness is constant in abatch. Thereaer, the wall is considered to be homogeneous. his not measurable with the experiment in elongational ow.Usually it is either accessible with uorescence microscopy ifpossible32 or with electronic microscopy33 aer a dehydrationstep of the capsule, which oen affects the dimensions of themembrane as compared to the wet state. Instead of measuring itwith classical methods, the previous experiments performed in

4566 | Soft Matter, 2014, 10, 4561–4568

the linear elastic regime allowed us to calculate it indirectly butin real suspension.

According to eqn (2) and (3), h is given by: h¼ 3Gs/E. In orderto obtain same size values for capsules to be associated withboth experiments, arbitrary interpolations on data points fromthe hydrodynamic experiments were determined (Fig. 9) withthe following regression: log(Gs) ¼ axb + c. The following values(a, b, c) values were obtained: capsules 5% HSA (�1.13 108,�3.799, �0.3372), capsules 10% HSA (�326.6, �0.8217, 6.579),capsules 20% HSA (�3.30 104, �2.194, 2.061). Consequently,the thickness of the membrane was calculated only for therange of capsule sizes that are covered by hydrodynamicexperiments.

The thickness increases with the capsule size (Fig. 11),notably in the case of batches 10 and 20% HSA. Moreover, at axed radius, the thickness grows with the concentration of HSA.These variations can be qualitatively understood, consideringthe capsule-processing. One limit case is the cross-linking of allmolecules of HSA in the drop at the interface. A simple massbalance shows that the thickness should increase linearly withthe capsule radius R and with the bulk HSA concentration C3D

HSA

in the case of a homogeneous membrane: h ¼ 3Gs/E ¼(R/3)(C3D

HSA/C2DHSA) where C2D

HSA is the concentration of HSA in themembrane. This relationship is qualitatively satised by ourresults, which combine deformation in an elongation ow andAFM studies. However, the chemical association at the interfaceis more complex. Aer the rst layer of HSA covers the interfacebetween oil and water, new HSA needs to cross over the layer toreach the terephthaloyl chloride to have the continuation of thecross-linking reaction. As the layer thickens, the transferbecomes limited by its permeability to HSA. Thus C2D

HSA is morean effective quantity, which takes into account a complexchemical process at the interface associated with a thickeningwhich could result in a non-homogeneous membrane. Adeeper understanding needs a better knowledge of kineticconstants, Marangoni-like and ow effects during interfacialpolymerization.

This journal is © The Royal Society of Chemistry 2014

Paper Soft Matter

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article Online

For the batch 20% HSA the wall accounts from 1.6% of theradius for a capsule of radius 64 mm to 9% for 120 mm. The sameincrease is observed for 10%HSA capsules, which reaches about11% for a capsule about 170 mm. A different behavior is pointedout for 5% HSA capsules, which have a constant ratio of about5%. This could question the validity of the assumption of aninnitely thin membrane in theoretical and numerical studies.

4 Conclusions

In this paper the capsule processing based on the interfacialcross-linking method is studied by investigating the mechanicalproperties of the capsule membranes produced at a givenemulsication rate. For each batch characterized by oneconcentration of HSA, the distribution of sizes has one or twopeaks with a typical range from 100 to 300 mm for 5% HSA forexample (Fig. 1). In each batch, the mechanical properties ofcapsules have been measured following two parameters: theinitial concentration of human serum albumin in the aqueoussolution and the radius of the capsule, which is a consequenceof the processing. Both surfacic and volumic Young's modulihave been extracted from small deformations (linear elasticregime) observed in two complementary experiments. The sur-facic shear modulus Gs, which is linear to the surfacic Young'smodulus, is determined by the extensions of individualcapsules plunged in an elongational ow with differentmagnitudes. The volumic Young's modulus E is deduced fromthe AFM force–indentation curves provided by local deforma-tions of themembrane in the linear elastic regime. Note that themethod based on an elongational ow has only been used onetime previously and on capsules of several millimeters of radiususing the famous four roll mills technique. Here, we extend theelongation method to capsules of typical radius of 100 mm witha microuidic-like chip.

As expected Gs and E depend largely on the proteinconcentration of the primary droplet. Surprisingly we observean enhanced behavior of the two elastic moduli with the size:typically 3 orders of magnitude for Gs for only a factor 3 in thecapsule size (10% HSA). This unexpected strong variation ispartly explained by the large variation of the volumic Young'smodulus characterizing the polymer network of themembrane measured by AFM. However this is not enough toexplain fully the variation of Gs. The other salient contribu-tion is due to the variation of the membrane thickness withconcentration and size. The thickness is deduced from thecombination of the values of the surfacic shear elasticmodulus and the volumic Young's modulus measured inde-pendently. For large sizes of capsules, Gs, E and h saturateabove 10% HSA rates. A material balance applied betweenmembrane HSA and volumic available HSA before polymeri-zation explains the qualitative variation with the size and HSAconcentration but is not quantitative to fully capture theunderstanding of the processing.

Nevertheless the relationship between size and mechanicalproperties gave highly sought information in terms of applica-tions. In the eld of chimioembolization the size and theresistance of the capsule have to be adjusted to the vessel to ll

This journal is © The Royal Society of Chemistry 2014

up. Moreover the data obtained should be used to optimize therelease kinetics of the encapsulated substances as in the eld ofcosmetic for which capsules, according to their size, should beused either to break under nger pressure or to deliver slowlytheir inner material through the intact membrane. As thecapsule size and mechanical properties are related, one way toreach the desired value of the capsule rigidity is to adjust itssize. Even if initial size distributions are pretty wide due to thepreparation process, various methods34 have been developed tosort particles in much narrower size distributions.

Experiments in elongational ow were conducted in theregime of small deformations in order to determine linearmechanical properties of microcapsules. Investigations in thelarge deformation regime should give valuable informationenabling the determination of the constitutive law of the HSAmembrane and its breakdown properties.

Acknowledgements

This work has beneted from the nancial support of CNES.This work has also been carried out in the framework of theANR CAPSHYDR (11-BS09-013-02), of the ANR polytransow(13-BS09-0015-01), of the Labex MEC (ANR-10-LABX-0092) andof the A*MIDEX project (ANR-11-IDEX-0001-02), funded by theInvestissements d'Avenir French Government programmanaged by the French National Research Agency (ANR).

References

1 A. Madene, M. Jacquot, J. Scher and S. Desobry, Int. J. FoodSci. Technol., 2006, 41, 1–21.

2 C. Onwulata, Annu. Rev. Food Sci. Technol., 2012, 3, 183–202.3 R. Fernandes and D. H. Gracias, Adv. Drug Delivery Rev., 2012,64, 1579–1589.

4 D. Demirgoz, T. O. Pangburn, K. P. Davis, S. Lee, F. S. Batesand E. Kokkoli, So Matter, 2009, 5, 2011–2019.

5 S. Banquet, E. Gomez, L. Nicol, F. Edwards-Levy, J.-P. Henry,R. Cao, D. Schapman, B. Dautreaux, F. Lallemand, F. Bauer,et al., Circulation, 2011, 124, 1059–1069.

6 R. K. Shah, J.-W. Kim, J. J. Agresti, D. A. Weitz and L.-Y. Chu,So Matter, 2008, 4, 2303–2309.

7 A. G. Skirtach, B. G. De Geest, A. Mamedov, A. A. Antipov,N. A. Kotov and G. B. Sukhorukov, J. Mater. Chem., 2007,17, 1050–1054.

8 S.-H. Hu, C.-H. Tsai, C.-F. Liao, D.-M. Liu and S.-Y. Chen,Langmuir, 2008, 24, 11811–11818.

9 Y.-C. Tan, K. Hettiarachchi, M. Siu, Y.-R. Pan and A. P. Lee,J. Am. Chem. Soc., 2006, 128, 5656–5658.

10 U. Seifert, Adv. Phys., 1997, 46, 13–137.11 D. E. Discher and F. Ahmed, Annu. Rev. Biomed. Eng., 2006, 8,

323–341.12 A. Fery and R. Weinkamer, Polymer, 2007, 48, 7221–7235.13 Z. Zhang, J. Stenson and C. Thomas, Adv. Chem. Eng., 2009,

37, 29–85.14 F. Risso and M. Carin, Phys. Rev. E: Stat., Nonlinear, So

Matter Phys., 2004, 69, 061601.15 M. Radmacher, Methods Cell Biol., 2002, 68, 67–90.

Soft Matter, 2014, 10, 4561–4568 | 4567

Soft Matter Paper

Publ

ishe

d on

12

May

201

4. D

ownl

oade

d by

Bib

lioth

eque

Int

erun

iver

sita

ire

DA

ix M

arse

ille

on

12/0

6/20

14 0

7:43

:27.

View Article Online

16 F. Dubreuil, N. Elsner and A. Fery, Eur. Phys. J. E: So MatterBiol. Phys., 2003, 12, 215–221.

17 R. M. Hochmuth, J. Biomech., 2000, 33, 15–22.18 I. Koleva and H. Rehage, So Matter, 2012, 8, 7672–7682.19 Y. Lefebvre, E. Leclerc, D. Barthes-Biesel, J. Walter and

F. Edwards-Levy, Phys. Fluids, 2008, 20, 123102.20 K. Chang andW. Olbricht, J. Fluid Mech., 1993, 250, 609–633.21 I. Koleva and H. Rehage, So Matter, 2012, 8, 3681–3693.22 D. Barthes-Biesel and J. M. Rallison, J. Fluid Mech., 1981, 113,

251–267.23 S. Ramanujan and C. Pozrikidis, J. Fluid Mech., 1998, 361,

117–143.24 C. Pozrikidis, J. Comput. Phys., 2001, 169, 250–301.25 E. Lac and D. Barthes-Biesel, Phys. Fluids, 2005, 17, 072105.26 W. D. Dodson III and P. Dimitrakopoulos, J. Fluid Mech.,

2009, 641, 263–296.

4568 | Soft Matter, 2014, 10, 4561–4568

27 K. Chang andW. Olbricht, J. Fluid Mech., 1993, 250, 587–608.28 M.-C. Andry, F. Edwards-Levy and M.-C. Levy, Int. J. Appl.

Pharm., 1996, 128, 197–202.29 T. Chu, A.-V. Salsac, E. Leclerc, D. Barthes-Biesel, H. Wurtz

and F. Edwards-Levy, J. Colloid Interface Sci., 2011, 355,81–88.

30 E. Lac, D. Barthes-Biesel, N. Pelekasis and J. Tsamopoulos,J. Fluid Mech., 2004, 516, 303–334.

31 S. Leick, A. Kemper and H. Rehage, So Matter, 2011, 7,6684–6694.

32 H. G. Bagaria, S. B. Kadali and M. S. Wong, Chem. Mater.,2010, 23, 301–308.

33 E. Brown, M. Kessler, N. Sottos and S. White,J. Microencapsulation, 2003, 20, 719–730.

34 A. Srivastav, T. Podgorski and G. Coupier, Microuid.Nanouid., 2012, 13, 697–701.

This journal is © The Royal Society of Chemistry 2014